Introduction

Salmonella Typhimurium is an enteric food-borne pathogen responsible for causing salmonellosis associated with acute diarrhea. The bacterial invasion initially requires flagellar motility to reach the intestinal lumen and to cross the mucus layer of the intestinal epithelia and subsequent adhesion to the host cell, mainly mediated by fimbriae1.

The key genes involved in the pathogenic process are encoded within highly conserved regions of the bacterial genome called Salmonella Pathogenicity Islands (SPIs). The key regulator HilA, which is positively regulated by HilC and HilD, is located in SPI-1; however, other virulence-related regulators, such as RtsA, are encoded outside this island2. During the invasion, proteins encoded by the prgHIJK, spaMNOPQRS, and invABCEFGH SPI-1 operons constitute a Type-3 Secretion System. Effector proteins, such as SipA and SipC (encoded in SPI-1) and SigD/SopB (encoded in SPI-5), translocate to the host cytosol through this system causing intracellular changes and inducing the immune response2,3,4. The survival and replication of intracellular Salmonella inside Salmonella-containing vacuoles (SCVs) is mainly mediated by the genes located in SPI-23. The immune response is activated through the recognition of pathogen-associated molecular patterns (PAMPs) by specific receptors expressed on the surface and/or inside different cell types1,5,6. The most important PAMPs are flagellin (monomers comprising the flagella filaments), particularly the FliC subunit, which is highly expressed on the bacterial surface, and lipopolysaccharide (LPS)7.

Oxygen availability is limited in the inflamed gut whereas other compounds, such as hydrogen sulfide (H2S)8 and H2 9 are abundantly produced by the colonic bacteria. In addition, nutrients are also limited inside the SCV, while reactive oxygen as well as nitrogen species may be found10. In anaerobic conditions, Salmonella is able to survive by using alternative energy sources, such as nitrate or fumarate, through the use of specific enzymes: nitrate and nitrite reductases11, fumarate reductase, DMSO reductase12,13 and respiratory hydrogenases9,14.

In a previous study, we compared transcriptomes of a clinical isolate of S. Typhimurium (strain 50-wt) and its derivative-multidrug-resistant mutant (strain 50–64) using microarrays15. In addition to showing antimicrobial resistance, 50–64 strain was also less invasive. In the present study, we focused on the genes of unknown function that showed impaired expression. For genes ybdJ, STM1441 and ynaJ a higher expression was seen in 50–64 compared to the wild-type strain (33.2-, 5.7- and 3.05-fold, respectively) whereas transcription was reduced in 50–64 for genes ydiY and yedF (−5.4- and −4-fold, respectively). We further selected those genes with a putative role in the acquired resistance phenotype or in the repressed virulence observed. Here, 4 putative membrane proteins and the yedF of unknown function were investigated to evaluate their involvement in these two phenotypes.

Results

Four genes potentially related to efflux (annotated as putative inner -ybdJ, STM1441 and ynaJ- and outer -ydiY- membrane proteins) were selected from a previous work for their differences at a transcriptional level15. In this work, an antibiotic resistant S. Typhimurium strain (50–64) showing in vitro impairment of the ability to be internalized in HeLa cells was compared to its susceptible counterpart (50-wt). Mutants of the reference strain SL1344 either carrying the disrupted genes or overexpressing them were obtained. Our experiments revealed that none of the genes were linked to antimicrobial susceptibility (data not shown) and statistically significant differences in the ability to invade HeLa cells were seen for both ∆STM1441 and the strain overexpressing this gene (STM1441_pBAD33). However, these results are contrary to what was expected as the poorly virulent mutant (50–64) showed higher transcriptional levels of STM1441 than 50-wt (Supplementary Figure 1). These inconclusive and contradictory results led to discontinuation of the study of these genes.

Another novel gene selected was the S. Typhimurium homolog of the E. coli yedF gene. Predicted to encode a TusA-like protein of unknown function, it has a 31% homology to the E. coli gene based on sequence alignment16. TusA is a sulfur transfer protein involved in tRNA modification and molybdenum cofactor biosynthesis in E. coli 16,17. From a structural point of view, the N-terminal region of yedF contains a CPxP conserved motif and the C-terminal domain has a similar folding structure to the translation initiation factor IF3C of E. coli. Together, these observations support the idea of a possible involvement of yedF in mRNA binding18,19. As seen for the other genes, neither mutants lacking yedF nor those overexpressing it showed any change in the antimicrobial susceptibility profile.

In vitro ability to interact with HeLa cells is compromised in the mutant lacking yedF

The deletion mutant lacking yedF (ΔyedF) showed a statistically significant 3.25-fold reduction in its ability to adhere/invade HeLa cells (p = 0.049) relative to the reference strain SL1344 (8.21% vs 26.64%, respectively). Furthermore, overexpression of yedF led to an increased ability to interact with the eukaryotic cells above the levels of the reference strain SL1344_pBAD33 (45.17% vs 33.40%, respectively), despite the difference not being statistically significant. These results were in line with the phenotype seen in 50–64 strain, suggesting the involvement of yedF in the virulence-associated phenotype.

Complementation of yedF in the ΔyedF background was not possible using the pBAD33 vector as the antibiotic used for the selection in both systems was chloramphenicol. For this reason, the gentamicin protection assay was repeated with a new collection of mutants using the p9817 vector conferring ampicillin resistance (Supplementary Table 1). Previously, the lack of expression of the gene yedF in the strains ΔyedF and its overexpression in the complemented strain ΔyedFp9817yedF (fold-change = 1642,44; pvalue < 0.01) compared with the wild-type strain SL1344 were confirmed by RT-PCR.

Then, the gentamicin protection assay was conducted and the results obtained were consistent with the previous findings; namely, the absence of yedF (for both ΔyedF and the mutant carrying the empty plasmid, ΔyedF_p9817) led to a statistically significant reduction (−3.24-fold and −2.62-fold, respectively) in the ability to interact with HeLa cells relative to the reference strain (8.21% and 10.16% vs 26.64%). When the gene was reintroduced, the capability of this strain to be internalized by epithelial cells was even greater than the reference strain achieving 40.32% (Fig. 1). This effect was likely due to the promoter used on the exogenous vector or to the copy number of the plasmid.

Figure 1
figure 1

Percentage of bacteria able to interact with HeLa cells for the reference strain (SL1344), mutant lacking yedF (∆yedF), mutant lacking yedF but carrying the empty plasmid (∆yedF_p9817) and the complemented strain (∆yedF _p9817yedF). The strain ∆yedF was obtained using the Datsenko & Wanner method by replacing the coding region of the gene yedF by a chloramphenicol-resistant cassette, and further transforming the empty vector p9817. The mutant lacking yedF was complemented in trans with p9817, a high copy number plasmid, carrying the gene of interest. The gentamicin protection assay performed consisted on infecting HeLa cells with the different strains (exponential cultures grown at 37 °C without shaking) at a multiplicity of infection of 100. The percentage of bacteria able to interact with HeLa cells was calculated as the ratio between the bacteria load used to infect eukaryotic cells and the amount of bacteria recovered at the end of the assay, being these results the mean of 3 independent experiments with standard deviation. One asterisk indicates a pvalue < 0.05.

Full size image

Transcriptomic analysis reveals that virulence-related genes are affected by yedF

A transcriptomic comparison between the SL1344 strain and ∆yedF, using the RNA-Seq, revealed that in the deletion strain 630 genes were repressed and 321 genes were overexpressed (p < 0.05). The complete list of genes with the corresponding fold-change values and a hierarchical cluster with the significant genes are reported as S1 Supplementary Data and Supplementary Figure 2, respectively.

In accordance with the results obtained from the gentamicin protection assay, a significant number of the repressed genes seen in ∆yedF were related to virulence. With respect to motility, the flagellum-related genes encoding for the late flagellin proteins FliC, the type III secretion apparatus FliR, and FlgA, involved in the flagellar assembly20,21 were repressed in ∆yedF relative to SL1344 (fold change of −2.17, −2.77 and −2.41, respectively) (Fig. 2 ). Indeed, electron microscopy experiments revealed that ∆yedF was aflagellated whereas SL1344 did show flagella on its surface (Fig. 3). However, this downregulation observed of fliC, together with the transcriptional repressor fljA and the phase 2 flagellin gene fljB (−11.48- and −5.97-fold, respectively) is contradictory. When flagellar phase variation is induced, FljA is expressed and inhibits the translation of the FliC mRNA by binding to its operator region. This results in the transcription of FljB and the absence of FliC22,23,24.

Figure 2
figure 2

Representation of the structure of the flagellum. This image represents the different components that constitute the flagellum (modified from the KEGG PATHWAY Database49). Red symbols indicate differentially expressed genes between the SL1344 strain and its mutant lacking yedF detected in the RNA-Seq assay.

Full size image
Figure 3
figure 3

Electron microscopy of the wild-type strain SL1344 and the mutant lacking yedF (∆yedF). Negative staining preparation of both samples was performed and visualization by Transmission Electron Microscopy (TEM) through a TEM JEOL 1010 equipped with a camera CCD Orius (GATAN) at an operating voltage of 80 kV. Images obtained revealed the absence of flagella in the mutant strain (B) whereas SL1344 showed intact flagella (A).

Full size image

In addition to motility, several chemotaxis genes required during the initial invasion2 were also down-regulated in ∆yedF. This was also the case for the type-1 fimbrial gene fimA and the regulatory gene fimZ (−5.1- and −3.07-fold, respectively).

Moreover, a broad repression of the genes encoded in the SPIs involved in the host invasion process (SPI-1, SPI2, SPI-4 and SPI-5) was also seen. Accordingly, significant repression was observed for the key regulators hilA and hilC (−3.10 and −3.57, respectively) in the yedF-deficient strain although repression was non-statistically significant for hilD (−1.89) (Supplementary Data).

To further investigate the involvement of yedF in virulence gene transcription, RT-PCR of fliC, fimA, hilA, hilD and invA was carried out in the ∆yedF strain complemented with p9817 (yedF) and compared to both the reference strain and the ∆yedF mutant. As expected, the expression levels of these virulence genes were restored, reaching even higher values than those observed for the reference strain SL1344_p9817 (Fig. 4). Surprisingly, 10-fold repression was observed for the fimbria-encoding gene fimA, which is significantly higher than repression seen in the knock-out strain (−1.45-fold). A possible explanation could be that yedF is a positive regulator of fimA transcription and excessive amounts of the regulator provoke a paradoxal opposite effect.

Figure 4
figure 4

Expression levels of the genes involved in virulence in the wild-type (SL1344), strain lacking yedF (∆yedF), mutant lacking yedF but carrying the empty plasmid (∆yedF_p9817) and the complemented strain (∆yedF_p9817yedF). RT-PCR of selected genes involved in virulence (fimA, fliC, hilA, hilD and invA) was performed for these four strains revealing that in all cases, the strains lacking yedF showed a reduction in the transcriptional levels compared with the reference strain (SL1344). When the wild-type gene was restored (as was the case in the complemented strain ∆yedF_p9817yedF) gene expression of all but one determinant (fimA) was increased above the levels of the SL1344. Transcription levels are represented as the fold-change calculated as the ratio between the value of relative quantitation of the mutants compared to the reference strain, SL1344, which value is 1 (mean of 3 independent experiments with standard deviations).

Full size image

Interestingly, genes involved in LPS biosynthesis, another relevant virulence determinant, were also repressed in ∆yedF (see Supplementary Data). LPS can be recognized by specific receptors of the host cells promoting innate inflammation pathways5,6. Transcription of the genes involved in O-antigen polysaccharide assembly (rfb genes) as well as in core-oligosaccharide composition (rfa genes)25,26 were repressed in ∆yedF (Fig. 5), with values ranging from a −2.03- to −6.79-fold change relative to SL1344. To validate the transcriptomic data, LPS content was measured. As expected, a decrease in the concentration of endotoxins was observed in the knock-out strain (∆yedF) compared to SL1344 (5.1 × 104 EU/mL vs 7.52 × 104 EU/mL). Similar results were seen when the endotoxin levels were checked in ∆yedF _p9817, the mutant carrying the empty plasmid (4.99 × 104 EU/mL). These levels increased to 6.46 × 105 EU/mL when yedF expression was restored by complementation, being all differences observed between the strains statistically significant (p = 0.05). Standard deviations for SL1344, ∆yedF, ∆yedF_p9817 and the complemented strain (∆yedF _p9817yedF) were 137.5, 115.4, 57.7 and 251.7, respectively.

Figure 5
figure 5

Representation of the LPS biosynthetic pathway. This image represents the different determinants involved in the biosynthesis of LPS (modified from the KEGG PATHWAY Database49). Components marked with red symbols indicate differentially expressed genes between the SL1344 strain and its mutant lacking yedF detected in the RNA-Seq assay.

Full size image

yedF is involved in transcriptional regulation of the anaerobic metabolism pathway

Transcriptomic analysis revealed that all genes overexpressed in ∆yedF relative to SL1344 encoded for proteins involved in anaerobic respiration (Table 1). In the absence of oxygen, Salmonella is able to survive by performing fermentations and/or by using alternative electron acceptors such as nitrate or fumarate13. Specifically, the genes overexpressed in the absence of yedF included: the cytoplasmic nitrite reductase NirB encoding genes (almost 40-fold increased expression), the nitrate reductase encoded by the narGHIJ operon (8- to 30-fold), the anaerobic DMSO reductase (almost 15-fold), and the nrdDG operon encoding for the class III reductase NrdD (12.5-fold). While the membrane-bound nitrate reductase NarGHI has shown to be responsible for NO production in NO− 3 conditions in Salmonella 27, the involvement of the cytoplasmic nitrite reductase NirB in NO production has only been previously demonstrated in E. coli 28 but not in S. Typhimurium27. Additionally, the genes encoding for the formate dehydrogenase H (FdhF), a component of the formate hydrogenlyase (FHL) complex, and the genes encoding for the maturation proteins HypABCDE that together catalyze the production of hydrogen from formate 28 were also activated in ∆yedF (2.67- to 10.38-fold).The transcriptional activator of the fdhF gene, named fhlA, was also up-regulated by 5.33-fold as was the glpABC operon, encoding for the anaerobic glycerol-3-phosphate dehydrogenase subunits (2.56- to 5.28-fold). In E. coli this enzyme has been reported to be essential for anaerobic growth on glycerol-3-phosphate29,30.

Table 1 Transcriptional values of representative genes overexpressed in the ∆yedF mutant compared to the wild-type SL1344. Fold change indicate the ratio between the expression levels in SL1344 and ∆yedF. For RNA-Seq, only statistically significant results (p < 0.05) are represented.
Full size table

The transcriptional up-regulation of all the genes involved in anaerobic metabolism seen in the present investigation is in accordance with a previous study conducted in E. coli 16. In that work, deletion of the sulfur transferase tusA triggered a similar situation to that observed here for hydrogenases and the already mentioned nar operon as well as nap operon in aerobic conditions. In the absence of tusA, the expression of the corresponding enzymes was increased suggesting that yedF, which was previously mentioned to be partially homologous to tusA, has a similar role in the regulation of these proteins in S. Typhimurium.

We further checked by RT-PCR the expression levels of the genes dmsA and frdA as representatives of the anaerobic metabolism, as well as the expression levels of yedF in the presence and absence of oxygen in the reference strain SL1344 (Fig. 6). Both anaerobic-related genes (dmsA, encoding the subunit A of the anaerobic dimethyl reductase, and frdA, part of the fumarate reductase operon), were statistically significant overexpressed in the absence of oxygen (10.42- and 4.50-fold change, respectively) compared to aerobic conditions. Moreover, a statistically significant repression of the expression of yedF was observed (−27.04-fold change) when oxygen was not present. These results are in accordance with the transcriptomic findings described in this section.

Figure 6
figure 6

Expression levels of dmsA and frdA, representative for the anaerobic metabolism, and yedF in the reference strain SL1344. RT-PCR of the genes dmsA, frdA and yedF was performed from the strain SL1344 obtained from exponential cultures in both aerobic and anaerobic conditions in three independent experiments. The results obtained indicate that an overexpression of the genes dmsA, encoding the subunit A of the anaerobic dimethyl reductase, and frdA, part of the fumarate reductase operon, occurred in the absence of oxygen, being these differences statistically significant. Moreover, the expression of yedF was seen to be repressed when oxygen was not available. These results indicate that under anaerobic conditions the expression of yedF is repressed whereas an activation of the expression of the anaerobic metabolism of Salmonella occurs. Transcription levels are represented as the fold-change calculated as the ratio between the value of relative quantitation of SL1344 in anaerobic conditions compared to the same strain in the presence of oxygen, which value is 1 (mean of 3 independent experiments with standard deviations). One asterisk indicates a pvalue = 0.05–0.01, and two asterisks a pvalue < 0.01.

Full size image

The production of proteins related to virulence and anaerobic metabolism is decreased in the absence of yedF

The proteomic approach revealed that 13 out of 16 proteins repressed in ΔyedF were related to pathogenesis: 6 were involved in chemotaxis, 3 were flagellar-related proteins (FlgE, FliG and FliC), 2 fimbrial proteins (FimA and FimC), and 2 were the invasion-related determinants SipC and SigD.

The remaining 3 proteins identified as being expressed at lower levels in ΔyedF relative to SL1344 belonged to the anaerobic metabolism. There were: i) the glycyl radical cofactor, produced only in the absence of oxygen by the class III ribonucleotide reductases (RNRs)31, ii) the ornithine decarboxylase (or SpeF) which buffers the extracellular environment under acidic conditions and has recently been demonstrated to act only in anoxic environments32 and iii) the formate dehydrogenase α-subunit, a member of the formate hydrogenlyase (FHL) complex, activated under anaerobic conditions33 (Table 2).

Table 2 Proteins with decreased production levels in the ∆yedF mutant compared to the wild-type strain SL1344.
Full size table

A partial correlation was found upon comparing the proteomic and transcriptomic analyses. All but 2 proteins identified as being expressed less in ΔyedF were also seen to be transcriptionally repressed; however, the differences were not statistically significant for half of them. The 2 proteins showing discordant results between the two techniques were the α-subunit of the formate dehydrogenase and the glycyl radical cofactor. Disparity between transcriptomic and proteomic data is often observed as there is no direct correlation between the levels of mRNA and protein expression.

Discussion

The studies of the four putative outer and inner membrane proteins have not led to conclusive results regarding their role in the phenotypes of antimicrobial resistance or virulence. This could be due to the fact that they play a minor role in efflux and an alteration of these proteins is not sufficient to see a relevant phenotype. Another explanation could be that they may be unrelated to virulence and/or antimicrobial resistance and are rather a collateral effect derived from the adaptation of the bacteria to antibiotic pressure.

More importantly, the present work revealed new insights into the role of the yedF gene and shed light on the different S. Typhimurium pathways in which this gene is potentially involved. Homology with the E. coli TusA suggests that YedF may be a TusA-like protein and that both genes could share at least some of the same functions19. In E. coli, TusA functions as a sulfur transferase delivering sulfur from the IscS protein to diverse molecules involved in a variety of pathways16,34,35. In addition, a previous study showed that an E. coli mutant lacking TusA presented severe growth defects36,37, suggesting its involvement in the general physiology of the bacteria. However, this function could not be attributed to yedF in S. Typhimurium as no growth alterations were detected in our ΔyedF mutant.

In our case, deletion of yedF triggered an increase in gene expression of enzymes involved in anaerobic respiration, similar to what has been demonstrated for tusA deletion mutant of E. coli 16. In this previous study on E. coli, yedF was also investigated and nitrate reductase activity was tested in both aerobic and anaerobic conditions using a strain lacking tusA, yedF and a triple mutant lacking these genes in addition to another TusA-like gene (yeeD). Neither yedF nor yeeD were able to replace, even partially, the functions of TusA16. However, according to our results, yedF seems to play a relevant role in the regulation of proteins involved in anaerobic metabolism in S. Typhimurium; although, another TusA-like gene with a protein sequence identity of 90% is also present in this bacterium (GI:486186069). Further studies using single and double mutants of both genes are needed to evaluate the contribution of both the tusA homolog and yedF in Salmonella.

Interestingly, we observed that all of the genes showing transcriptional activation in ΔyedF mutant are regulated by FNR, the main transcriptional regulator of the adaptive response to the lack of oxygen in S. enterica 9,13,16,27,29,30,33,38. However, this transcriptional regulator was not affected by the presence or absence of yedF. We found that in the absence of yedF, even when oxygen is available and most of the FNR is inactive39, a significant activation of the genes encoding for the anaerobic enzymes occurs, suggesting that yedF contributes to the regulation of these enzymes. This was not seen at a protein level, as only the hydrogenase HypB and the fumarate reductase FrdA were detected but did not show differential amounts between the strains studied. These results are expected as protein synthesis of enzymes involved in the anaerobic metabolism is not needed in aerobic conditions.

Dahl et al.16 proposed a model in which interaction of the IscS protein with TusA decreases the pool of available IscS needed for FeS cluster biogenesis and, consequently, for the activation of FNR. As YedF also contains the same conserved cysteine, the catalytic residue involved in the interaction with IscS39, we propose that this may have occurred in the present study. Based on the available information and the results obtained in this work, we propose that yedF affects gene activation by influencing FeS cluster biosynthesis in the cell, in the same way as the model previously proposed for TusA16. Supporting this statement, in this work we also show that in the wild-type strain SL1344, when oxygen is not available, yedF is repressed whereas the anaerobic metabolism-related genes dmsA and frdA are overexpressed compared to aerobic conditions.

In addition, we propose that yedF may have a role in virulence of Salmonella. Our results demonstrate a decrease in virulence-related determinants at both gene and protein levels, which correlated with the reduced in vitro ability to interact with epithelial cells of the strain lacking yedF and was restored when the gene was reintroduced. Virulence regulation is a complex process with many contributing factors leading to gene expression in particular conditions; therefore, the interpretation of the specific role of yedF in this system requires further study. Nevertheless, our results indicate that this gene is involved in the regulation of virulence as the absence of yedF results in significant changes. Accordingly, we propose this gene to be named vrf, for virulence-related factor. Taking into account this multifactorial regulation, an alteration of the signaling network could, in some cases, lead to paradoxical situations, such as the unexpected gene expression profile of determinants involved in the flagellin biosynthesis pathway.

In addition, we found that LPS was also affected by yedF. Previously, Kong et al.40 reported that alteration of the LPS composition in S. Typhimurium led to a decrease in virulence in an in vivo murine model. In that study, mutants obtained by the deletion of genes involved in the O-antigen and core biosynthesis of LPS, also shown to be transcriptionally repressed in the present work, were administered orally in mice. The study revealed that intact LPS was required for optimal invasion and colonization of host tissues. Accordingly, the effect of yedF on LPS composition may partially contribute to the decreased ability to interact with HeLa cells detected in the yedF-defective mutant in addition to the proper repression of the invasion genes.

In the present study, we have identified a transcriptional regulator, yedF, which activates many genes involved in the host-pathogen interaction. Thus, we propose that yedF contributes to the activation of the virulence machinery. Taking into account the pleiotropic functions of yedF observed in this work, we hypothesize that yedF is activated in the initial steps of infection and contributes to positive transcription of the genes involved in motility and attachment, facilitating the arrival of bacterium to the host cell. In addition, yedF may be involved in the activation of genes encoded in SPIs needed for its internalization. When oxygen is available, yedF contributes to keeping the anaerobic machinery repressed; while when bacteria encounter conditions in which oxygen is absent, such as in the intracellular environment, transcription of yedF is repressed, allowing the activation of the anaerobic metabolism pathway. Nevertheless, further characterization of this factor needs to be undertaken in order to validate this theory.

Methods

Clinical isolates and genes selected

In a previous work published by Fàbrega et al.15, a S. Typhimurium clinical isolate and an antibiotic-resistant mutant, 50–64 was obtained. Five unknown genes potentially involved in antibiotic resistance/virulence were selected based on previously performed microarray analyses (>2-fold significant difference)15. These genes included three putative inner membrane proteins: STM1441 (Gene ID: 1252959), ynaJ (Gene ID: 1253180) and ybdJ (Gene ID: 1252102); one putative outer membrane protein named ydiY (Gene ID: 1252845) and one hypothetical transcription factor, yedF (Gene ID: 1253487). Sequences were obtained from the S. Typhimurium LT2 reference strain (RefSeq NC_003197.1).

Construction of ΔSTM1441, ΔynaJ, ΔybdJ, ΔydiY and ΔyedF mutants

Individual inactivation of the genes of interest was done in the reference strain SL1344 following the Datsenko & Wanner method41. Strains and primers used are listed in Supplementary Table 1 and Supplementary Table 2, respectively. Selection of the knock-out colonies was made by antibiotic selection using Luria-Bertani (LB) supplemented with 8 mg/L chloramphenicol (Sigma-Aldrich) and 10 mM arabinose (Panreac). Confirmation of the deleted region was done by PCR, and the absence of gene expression of the genes was checked by RT-PCR using specific primers (Supplementary Table 3).

Obtainment of strains overexpressing STM1441, ynaJ, ybdJ, ydiY and yedF

Constructions of the cloning vector pBAD33 carrying the full genes STM1441, ynaJ, ybdJ, ydiY and yedF were made using their respective primers (Supplementary Table 2). These fragments, as well as the vector pBAD33, were purified and digested with the restriction enzymes XbaI and SacI. The primers used for amplification were pBAD_F (5′-CTGTTTCTCCATACCCGTT-3′) and pBAD_R (5′-CTCATCCGCCAAAACAG-3′) published by Guzman et al.42. The resulting mutants (Supplementary Table 1) were selected in LB containing 30 mg/L of chloramphenicol (Sigma-Aldrich). Gene expression was induced with 10 mM arabinose (Panreac), and overexpression was confirmed by RT-PCR using the primers shown in Supplementary Table 3.

Complementation of the expression of yedF in the ΔyedF mutant

The ∆yedF strain with a disrupted yedF was complemented in trans with p9817, a high copy number plasmid, carrying the gene of interest (∆yedF_p9817yedF). Both plasmid p9817 and the amplified fragment of the gene of interest containing the restriction enzyme sites were digested with NdeI and BamHI. Confirmation of the recombinant plasmid containing the gene of interest was made by PCR and DNA sequencing using the primers 1681.for (5′-CCCCAGGCTTTACACTTTATGCTTCC-3′) and 1030.rev (5′-GCGGATGCCGGGAGCAGACAAGCCC-3′) at an annealing temperature of 57 °C. Positive colonies were selected in LB containing 50 mg/L of ampicillin (Sigma-Aldrich). The strains obtained in this study are listed in Supplementary Table 1.

Antimicrobial susceptibility testing

The MICs of nalidixic acid, ciprofloxacin, tetracycline, cefoxitin, erythromycin and trimethoprim were determined by Etests (Biomérieux) on Mueller Hinton-II plates (Becton Dickinson) following the manufacturer’s recommendations. Three replicates of each susceptibility test were performed.

Gene expression by RT-PCR

RNA of the strains studied was obtained from exponential cultures, in the presence and absence of oxygen, when required, using the Maxwell 16 Research Instrument (Promega) and the Maxwell 16 LEV simplyRNA Blood Kit according to the manufacturer’s recommendations. Then, a two-step RT-PCR was performed in a StepOne Real-Time PCR System (Applied Biosystems) as previously described by our lab43. The primers used are shown in Supplementary Table 3. Three independent extractions and analyses were performed.

Bacterial growth

OD at 620 nm of overnight cultures was determined by means of an iEMS Multiskan Reader MF (Thermo Fisher Scientific) as described previously15. Each plate included four replicates of each sample, and the assay was repeated three times.

Gentamicin protection assay

The in vitro assay by bacterial infection of HeLa cells was performed according to Fàbrega et al.15. When needed, induction of gene expression was done with 10 mM arabinose.

Endotoxin detection assay

Measurement of the endotoxin content of SL1344_p9817, ∆yedF_p9817 and ∆yedF_p9817yedF was performed in triplicate by means of the kinetic-chromogenic LAL Kinetic-QCL kit (Lonza).

RNA-Seq

RNA of SL1344 strain and ∆yedF was extracted in triplicate as described previously in Methods. Then, RNA was quantified using a Quantus Fluorometer (Promega) and integrity was assessed with a 2100 Bioanalyzer (Agilent Technologies). Ribosomal RNA was depleted with the Ribo-Zero Magnetic Kit for Gram-negative bacteria (Epicentre). Then, libraries were generated with the TruSeq Stranded mRNA Sample Prep Kit (Illumina) and sequenced with the Illumina MiSeq platform (2 × 75 bp). An average of 4 million reads/sample was obtained (average Phred quality score = 37). Reads were mapped onto the reference genome (SL1344 NC_016810.1) using EDGEpro software44. The resulting count datasets were exported to the DESeq. 2 module in R45 for normalization, and pair-wise differential expression was carried out for each gene. Only ± 2 fold change values between both conditions were considered (pvalue < 0.05). Gene Ontology and Pathway analysis according to the KEGG database were conducted with DAVID46,47 for all significant genes. Hierarchical clustering of Euclidean distances of the differentially expressed genes was obtained using the pheatmap package in R. Transcriptomic data has been uploaded in the Gene Expression Omnibus (GEO NCBI) Platform (Accession reference “GSE101075”).

Negative staining electron microscopy

For Transmission Electron Microscopy (TEM) samples of the strains SL1344 and ∆yedF were prepared from an overnight culture grown aerobically. Pellets were washed twice with PBS, resuspended in 1 mL of a phosphate buffered solution 0.1 M and pH = 7.5 containing 2.5% of glutaraldehyde and kept at 4 °C. Then, a drop of the sample was deposited on a formvar coated copper grid and after 20 min, the excess was removed and a drop of 2% uranyl acetate was deposited for 1 min. Once the grid was dried, observation was done through a TEM JEOL 1010 equipped with a camera CCD Orius (GATAN) at an operating voltage of 80 kV.

Comparative proteomic analysis iTRAQ

Differences in protein abundance between the SL1344 and ∆yedF strains were performed using the isobaric tag for relative and absolute quantitation (iTRAQ) technology. Four independent replicas were prepared per strain. Exponential cultures were obtained and cell pellets were finally resuspended in 8 M urea, 2 M thiourea 2.5% 3-[(3-cholamidopropyl) dimethylammonio]-1 propanesulfonate (CHAPS), 2% ASB-14, 40 mM Tris-HCl, pH 8.8. Then, cells were sonicated and the extracted proteins were trypsin digested, and peptides were labeled with 8-plex iTRAQ reagent (ABSciex). Peptides were analyzed by liquid chromatography coupled to mass spectrometry (LC-MS/MS) using a nanoAcquity (Waters) coupled to LTQ-Orbitrap Velos (Thermo Scientific). The Proteome Discoverer software (v1.4) was employed for database search and reporter ion intensities extraction using SequestHT search engine. Further analysis (normalization, protein ration calculation and statistical analysis) were carried out using R Development Core Team/ Inferno RND software. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD006825.

Statistical analysis

Non parametrical statistical test consisting in a Kruskal-Wallis rank sum test was performed with R 3.3.343,48. P values ≤ 0.05 were considered as significant.